MECHANICAL PROPERTIES FORECAST IN COMPOSITES USING NEURAL NETWORKS
S. Sánchez-Caballero, José Enrique Crespo Amorós, Francisco José Parres, M.A. Selles, Rafael PLA-FERRANDO · ANNALS OF THE ORADEA UNIVERSITY Fascicle of Management and Technological Engineering · 2011
Abstract: The aim of this paper is to introduce a method to forecast the mechanical properties of a composite based on its constitutive materials using a neural network. As input data, a limited number of tests to train the network are needed. From them it will be possible to make forecasts, with a less than 1 % of error, the material properties. The forecasts can be done, not only inside the training range but also outside but with an unbounded error rate. 1. INTRODUCTION. The use of technical plastics has been extensive for the last years in almost all industrial sectors. The main drawback lies in the materials price. The continual increase in price of commodities and the increase of competition pull the companies to reduce its